Human Whole blood transcriptome 10 or 30 Million reads

Question

[HMISEQ]

Hello All,

It is my first mail to this community and I am new to Illumina RNA-seq method. I have previous experience in 454 amplicon hi-throughput methods. We have human whole blood samples collected in pax-gene tubes from a clinical trial and we are interested to investigate gene expression profiles (not interested in novel transcripts, variants, snps) between the vaccinated and non-vaccinated individuals and their response during infection. I have few questions:

Do globin reduction during initial library preparation will be really useful?

I have gone through some publications and but not sure whether to go for single end or paired end reads?

Shall we go for Mi-seq (10 million reads/per sample) to reduce the cost of the runs instead of 30-35 Millions reads per sample in Hi-seq instrument?

Thanks in advance for your replies.

[Devon Ryan]

Moving to questions, anyone disagreeing can feel free to move it back to the forum section.

Answer 1

[Devon Ryan]

This is probably more appropriate for SEQanswers.

Do globin reduction during initial library preparation will be really useful?

While I've not done RNAseq from blood, I would be very surprised if globin reduction isn't helpful. Since globin transcripts will be a huge portion of your extracted RNA and you're not interested in it, not depleting it will just increase the amount of sequencing you'll need to do.

I have gone through some publications and but not sure whether to go for single end or paired end reads?

For standard DE like you're doing, single-end is sufficient.

Shall we go for Mi-seq (10 million reads/per sample) to reduce the cost of the runs instead of 30-35 Millions reads per sample in Hi-seq instrument?

10 million reads/sample is normally sufficient (in fact, you'll see that number mentioned in papers looking at power as a function of read number). Note that it might still be cheaper to just multiplex things on a HiSeq, depending on the number of samples.

[User 59]

Indeed, with no mention of numbers of samples, I can't see many instances where a MiSeq would be cheaper for this application.

[HMISEQ]

Devon-Thank you for the answers and useful info.

Daniel-We will end up with atleast 150 samples in total.

[User 59]

You'll definitely want to hit the HiSeq then. You do realise you can pick your read numbers in the HiSeq, it just depends how much you want to multiplex the lanes. 12 samples/lane I think is a good price point for gene level differential expression studies.